Speed up Tile painting on large meshes

{
  "backend": {
    "author": {
      "email": "drschwabe@fastmail.com",
      "name": "Derrick Schwabe",
      "time": 1790305341
    },
    "commit": "git:sha1:65b11bce3f97b6ce30c07bd722860fe23e71ffcc",
    "committer": {
      "email": "drschwabe@fastmail.com",
      "name": "Derrick Schwabe",
      "time": 1790305341
    },
    "manifest": "mmm-manifest:sha256:28a0b35863a10c9e4581c5e87f38daefaf5ba3d7298b8abef2ba3d2adc6c58c7",
    "message": "Speed up Tile painting on large meshes\n\nEach dab awaited a full mesh render before the next could stamp, and every\nrender redrew every Tiled canvas. Dabs now go into state as the pointer\ncrosses cells, with one render in flight per stroke and at most one more\nqueued. Map, tileset and stamp canvases record the inputs they were drawn\nfrom, and the per-render pass skips those that are unchanged. Source images\nare found once per render rather than scanned for per baked node.\n\nOn a 771-node mesh with 557 stamps, together with the two previous commits,\na single stamp now renders in about 1.35s instead of 3.1s, and a 16-tile\nstroke settles in about 3s instead of 7.7s.",
    "objectFormat": "sha1",
    "parents": [
      "git:sha1:01a87fc384a95483e95b1fd290fa4f41730db30b"
    ],
    "ref": "refs/heads/master",
    "summary": "Speed up Tile painting on large meshes",
    "tree": "git:sha1:f2a37ff0196f089057c582e679fe93e1ea5044c2",
    "type": "git"
  },
  "issuedAt": "2026-09-28T15:47:36.878Z",
  "previous": "mmm-revision:sha256:230e9b33884b9d8640f1c22bb7166b499845050f1a17def3f7fe87dc47242a4b",
  "project": "mmm1np06566gf6u8fpyj84u2emf25l98qad36700d2nlep6jzqdknp8qkyd3xu",
  "protocol": "mmm/0-alpha",
  "sequence": 255,
  "type": "mmm/revision"
}